A Data Mining Approach for Retailing Bank Customer Attrition AnalysisAiyappa N D 1NT13CS010Aniruddha Achar 1NT13CS016Akshaj B 1NT13CS032

Introduction

Deregulation within the financial service industries and the widespread acceptance of new technologies is increasing competition in the finance marketplace.

In the financial area, data mining has been applied successfully in determining various aspects.

The goal of attrition analysis is to identify a group of customers who have a high probability to attrite, and then the company can conduct marketing campaigns to change the behavior in the desired direction.

Process of data mining for bank attrition analysis

Problem definition:

formulation of business

problems in the area of customer

retention.

Data review and initial selection.

Problem formulation in

terms of existing data.

Data gathering, cataloging and

formatting.

Data Processing: (a) Data

cleansing, data unfolding and time-sensitive

variable definition, target

variable definition, (b)

Statistical analysis, (c) Sensitivity

analysis, (d) Leaker detection,

I Feature selection.

Data modeling via classification models: Decision

Trees, Neural Networks, Boosted Bayesian Networks, Selective Bayesian

Network, an ensemble of classifiers.

Result review and analysis: use the data mining model to predict the likely attriters

among the current

customers.

Result Deployment:

target the likely attriters (called

rollout.)

Business Problem

The client was one of the top 10 retailing banks in the world. It offered many types of financial retail products to various customers. The product discussed in this paper belongs to a certain type of loan service.

There are three types of attriters High attrition rate Slow attriters Pirated accounts

Problem Definition

Slow attriters: Customers who slowly pay down their outstanding balance until they become inactive. Attrition here is understood comprehensively, where voluntary attrition can show more than one behavior.

Fast attriters: Customers who quickly pay down their balance and either lapse it or close it via phone call or write in.

Cross selling: Identify customers who are likely to purchase alternative products offered to existing loan customers such as life insurance and the like. The increase in relationships is believed to serve as a deterrent to attrition.

High risk: Customers who are likely to become high risk. Pirating: Identify customers likely to transfer their relationship to competing products

and away from our client.

User state model

Problem classes

Problem Class #1: Retention of Existing Customers The problem requires the stratification of customer segments by leveraging current segmentation model in order to: Develop models that predict the customers who are likely to attrite within 30 to 60 days on an

ongoing basis. Identify the characteristics of the most profitable/desirable customer segments in order to

develop policies to ensure their continued support, to grow the group, and to acquire more customers with similar characteristics.  

Problem Class #2: Customer Activation Policies Identify customer groups whose characteristics lend them to migrating from unprofitable/dormant to profitable. Once identified, the characteristics can enable the development of risk, maintenance and opportunity policies tailored to a successful migration.

Data Selection

Data Preprocessing Goals Data Period Identification Time Series “Unrolling” Target Value Definition First Stage Statistical Analysis Data Pre-modeling Field Sensitivity Analysis and Field Reduction Files Set Generation

Data Mining Model Development Process

Evaluation Criterion: Lift Prediction accuracy, which was used to evaluate the mining algorithms, is not a suitable

evaluation criterion for the data mining applications such as attrition analysis Data Mining Models Based on Different Algorithms

Boosted Naive Bayesian Neural Network Decision Tree Selective Naive Bayesian (SNB)

Boosted Naive Bayesian Networks (BNB)

Decision Trees

Neural Networks

Selective Naive Bayesian Networks

Data Mining Findings

Field Test

The top percentage of the customer attrition list does contain concentrated attriters The data mining based marketing approach is effective for retention purpose. They ran the model generated from the ensemble of classifiers approach on the

current customers and then sorted the customers based on the attrition scores.

Conclusion

In this paper, a data mining approach for retailing bank customer attrition analysis was presented.

We discuss the challenging issues such as highly skewed data, time series data unrolling, leaker field detection etc., and procedure of a data mining task for the attrition analysis for retailing bank.